Jacob Munson (Dept. of Mathematical Sciences, MSU) 

9/28/2023  3:10pm

Abstract:

Recommender systems have become integral to our online experiences, offering personalized content recommendations that shape our digital interactions. However, despite their widespread use, these systems often operate as a black box, providing recommendations with little insight into the reliability of their suggestions. This opacity raises important concerns, especially when the stakes are high, such as in medical diagnosis or financial investments. In this talk, we review the relatively underdeveloped field of uncertainty estimation within recommender systems, aiming to demystify the black box nature of explicit feedback collaborative filtering. We explore a range of state-of-the-art empirical uncertainty measures and their practical applications in Top-N recommendations, while also critically evaluating their quality and limitations.